Presenter: Andreas Tjarnberg, Postdoctoral Associate, New York University
Title: Translating gene regulatory network inference into a white-box deep learning framework
Abstract: Gene regulatory network (GRN) inference is made difficult due to noise and the scope of possible explanatory models. The deep learning framework in general has shown great results for a number of tasks in multiple fields dealing with noisy and complex data. However, one of the major hurdles for adapting the deep learning framework to GRN inference is that the end goal of GRN inference requires interpretable models. To this end, I will present my attempts at white-box modeling and adapting the current state-of-the-art GRN inference methods (the inferelator) to the deep learning framework in order to develop an interpretable model that can merge the deep learning framework with the goals of GRN inference.
Secondly, I will introduce my attempts at formulating a cell local and gene-specific rate of change estimate (velocity) for single-cell data in absence of RNA velocity. Here I compare RNA velocity and local rate of change estimates and discuss how this may be used for the task of GRN inference.